Total Survey Error in Practice

6Mixing Modes: Tradeoffs Among Coverage, Nonresponse, and Measurement Error

Roger Tourangeau

Westat, Rockville, MD, USA

6.1 Introduction

As de Leeuw (2005) points out, there are a number of reasons to use mixed‐mode designs in surveys. Among the more common motives are to reduce data collection costs, improve coverage of the target population, increase response rates, or reduce measurement errors. This chapter defines mixed‐mode surveys as surveys that use multiple modes of data collection; some authors (Schouten et al., 2013) broaden this definition to include surveys that use methods to contact or recruit respondents that are different from the method used to collect the data. Unfortunately, in the U.S.A. at least, this broader definition includes most face‐to‐face and telephone surveys, since such surveys typically begin with an advance letter mailed to sample members (wherever an address is available for them); this blurs the distinction between surveys that combine data obtained via different methods with those that use only a single data collection mode. Although the method of recruitment can affect the coverage and nonresponse properties of the survey, in this chapter, we focus on surveys that use multiple modes to collect data.

The appeal of mixed‐mode surveys has increased as response rates have continued to fall in the U.S.A. and elsewhere (Brick and Williams, 2013; de Leeuw and de Heer, 2002; Tourangeau and Plewes, 2013) and as survey costs have continued to rise ...